Landing pages that convert: what the evidence supports

Most conversion advice is folklore repeated with confidence. A smaller set of things reliably works, and nearly all of it is about reducing the effort a visitor has to make.

A landing page has one job: let someone who arrived with a question decide, quickly, whether you are the answer. Nearly every reliable improvement comes from removing something that got in the way of that.

The structure that holds up

• Say the same thing the ad said. Message match is the most common failure. If the ad promised a price and the page opens with a brand story, the visitor has to translate, and a share of them will not.
• Lead with the specific claim. What it is, who it is for, what it costs or how it is priced. Vague headlines protect nobody.
• Show evidence early. Numbers, named clients, a screenshot, a real photograph of the actual thing. Stock imagery signals nothing.
• One primary action, repeated down the page. Secondary actions are fine as long as they look secondary.
• Answer the objection on the page. Price, contract length, timelines, who owns what. Every unanswered objection becomes an abandoned session.

Friction worth removing

Every extra form field costs completions. Ask for what you need to make the first call and nothing more; you can qualify in the conversation. Slow pages cost completions in proportion to the delay, and mobile is where the losses concentrate. Interstitials, chat widgets that open unprompted, and cookie banners covering the call to action are all self-inflicted.

A mundane check that repeatedly finds money: complete your own form on a mid-range phone, on mobile data, as a new visitor. Time it. Note every moment you had to think. That list is usually the roadmap.

What testing can and cannot tell you

A/B testing needs traffic and patience. At a few hundred visits a week you cannot detect anything short of a large effect, and stopping a test the moment it looks positive manufactures results that do not repeat. Below roughly a thousand conversions a month, spend the effort on obvious fixes and qualitative evidence — session recordings, five customer interviews, the questions your sales team keeps answering — rather than on a testing programme you lack the volume to run.

When you do test: one meaningful change at a time, a duration fixed in advance, whole weeks to avoid day-of-week effects, and a metric that is a business outcome rather than a click.

The button colour question

It does not matter, beyond the button being visible and obviously clickable. Button-colour case studies persist because they are cheap to run and easy to write up. The tests that produce durable gains are almost always about the offer, the proof or the effort required — the parts that need a decision from someone senior.

Before you redesign

Redesigns reset what you learned and frequently lose ground that was invisible until it was gone. Fix the specific failures first, measure, and redesign only when the page is structurally wrong rather than merely dated. If you are redesigning anyway, read how to do it without losing rankings, and put the measurement in place before the change: see GA4 that survives scrutiny.